TU Wien:Sicherheit, Privacy und Erklärbarkeit in Maschinellem Lernen VU (Rauber, Mayer)

Aus VoWi
Zur Navigation springen Zur Suche springen


Daten[Bearbeiten | Quelltext bearbeiten]

Diese LVA wird nicht mehr von dieser Person angeboten, ist ausgelaufen, oder läuft aus und befindet sich daher nur noch zu historischen Zwecken im VoWi.
Vortragende Rauber Andreas , Mayer Rudolf
ECTS 3
Sprache English
Links tiss:194055 , Homepage
Zuordnungen
Masterstudium Data Science
Masterstudium Business Informatics
Masterstudium Software Engineering & Internet Computing
Masterstudium Logic and Computation

Mattermost: Channel "sicherheit-privacy-und-erklaerbarkeit-in-maschinellem-lernen"RegisterMattermost-Infos

Inhalt[Bearbeiten | Quelltext bearbeiten]

noch offen, bitte nicht von TISS/u:find oder Homepage kopieren, sondern aus Studierendensicht beschreiben.

Ablauf[Bearbeiten | Quelltext bearbeiten]

noch offen

Benötigte/Empfehlenswerte Vorkenntnisse[Bearbeiten | Quelltext bearbeiten]

It is assumed that you have taken 184.702 Machine Learning.

Vortrag[Bearbeiten | Quelltext bearbeiten]

Guter Vortrag, insbesondere Prof. Rauber merkt man die Begeisterung für das Thema an.

Übungen[Bearbeiten | Quelltext bearbeiten]

SS22:

  • There are two exercises which were supposed to be done in pairs.
  • The 1st exercise was about investigating different approaches to Explainability, where you got 2 models, a test dataset and a few data instances and had to explain and compare the decision boundaries with tools like for example PDP, PyALE, LIME, Shap, etc. First, each student had to do an analysis individually, but there were certain tasks where you were supposed to collaborate and exchange your findings. You were not allowed to do any model reverse engineering. Besides delivering your code, you also had to write a report where you answer questions from the assignment description and explain your findings. The questions were often not that clearly formulated and you had the feeling of repeating yourself for answering them. The 1st assignment was announced on 20 March 2022 and was due on 24 April 2022.
  • The 2nd exercise is a project where you can chose from a list of topics, including Privacy-preserving data publishing, Privacy-preserving computation and Adversarial Machine Learning. The 2nd exercise came much later than announced, had to be done in summer (July). The topic registration for your group was available around 13 June 2022. You first had to submit a draft concept (1–2 pages) document for your project, including a description of your topic and chosen solution. The 2nd exercise was due end of July 2022.

Prüfung, Benotung[Bearbeiten | Quelltext bearbeiten]

noch offen

Dauer der Zeugnisausstellung[Bearbeiten | Quelltext bearbeiten]

noch offen

Zeitaufwand[Bearbeiten | Quelltext bearbeiten]

You should plan at least 3 days of concentrated study, the slide sets are several hundred pages, build on understanding and this is also tested in the exam.

I think that this LVA is more than 3 ECTS effort, as both exercise are a lot of work and the exam content consists of around 900 slides.

Unterlagen[Bearbeiten | Quelltext bearbeiten]

noch offen

Tipps[Bearbeiten | Quelltext bearbeiten]

noch offen

Verbesserungsvorschläge / Kritik[Bearbeiten | Quelltext bearbeiten]

noch offen